jump to navigation

The Printed Transistor: why Chris is right and Dave is wrong January 10, 2011

Posted by mareserinitatis in engineering.
Tags: , ,
trackback

I’ve been getting caught up on the Amp Hour, and an argument that seems to keep coming up is whether there will be anything like an ‘in-house’ fab where people can manufacture their own ICs. Chris thinks it’s coming, but Dave insists it’s not.

First, there are a lot of good reasons for wanting to be able to manufacture your own ICs: •you don’t have to wait for the part
•even if ICs are cheap, you have to buy several which you may or may not need
•some day, shipping will be going up (when we run out of oil – or maybe before then)

I think Chris will end up winning this argument in the long run. Chris keeps pointing to 3D printers as an example of cool manufacturing technology that people never thought would be affordable. It actually turns out that this is exactly why he’s right.

Dave keeps saying there’s no way you could put a ‘fab’ inside someone’s house. It’s dangerous, nasty, expensive, etc. And he’s right. However, that’s not how ICs of the future are necessarily going to be made. (Okay, probably they will be made that way in bulk, but not individually.)

Now that 3D printing has become viable, a lot of research is investigating the possibility of using it to prototype electronics. Inks for these printers are being developed that could be used to print resistors and dielectrics for caps.

However, there are also people developing processes to print transistors. The scale is larger than you’d find in a fab-manufactured ICs. Right now, they can make transistors out of thin films and polymers using printing or similar deposition techniques. There are certainly issues with the processes, reliability, etc. However, it isn’t unreasonable to think that five or ten years down the line, 3D printers will very easily be able to use these deposition processes to create transistors and ICs right in your home lab. Even with the ability, however, I don’t imagine IC engineers ought to be looking for new jobs quite yet.

High-Performance All-Inkjet-Printed Transistors for Ultra-low-cost RFID Applications

The huge impact of printed transistors

Printed and Thin Film Transistors and Memory 2008-2028

Advertisements

Comments»

1. Charles J Gervasi - January 10, 2011

I have worked at two companies that owned milling machines to make PCBs. The machine mills away copper, the way etch would, and leaves traces. It drills through-holes, which you must run a wire through if you want them to act as vias.

I have heard of people similarly making their own boards through an etch process, but I’ve never seen it done.

I thought maybe this technology would improve making it easier to run boards in the lab. It didn’t happen.

Instead McFabhouses cropped up that batch orders making it easy to order a few boards very cheaply.

Question: Why won’t the same thing happen with IC fabs?

Is it that important to be able to do it on-site in a small lab? If you can send your files off to a fab and get an IC back quickly, isn’t that really what we’re after.

Every time this topic comes up on The Amp Hour I think Chris is right that inexpensive, quick-turn, customized ICs are coming. I also think Dave is right that it won’t happen in your basement or a small lab.

Like

2. mareserinitatis - January 10, 2011

Actually, where I work, we have two PCB mills and a 3D printer. We use this stuff for prototyping, making antennas, etc. Granted, we’re not a small lab…but we’re not huge, either. Maybe it’s somewhat dependent on what you’re doing. For research, especially in places like where we’re at where there aren’t these huge government-funded labs (i.e. smaller universities or smaller companies), I can see that they’d be useful…or if you’re running your business out of your home.

Like

3. Dave Jones - January 10, 2011

You’ve missed all my points about why it won’t become generically popular like 3D printing, even if a machine that can spit out chips may eventually happen in some way shape or form.
But such a machine would only ever be incredibly niche, unlike 3D printers, and that’s the crux of it.
I won’t repeat the reasons here, they are on the AmpHour comments sections.

Dave.

Like

mareserinitatis - January 10, 2011

Your argument is that it’s cheaper and easier to get chips that have been mass produced. However, this is like Charles’ argument: why have a PCB mill when you can order boards cheaply from some boardhouse?

As people have mentioned on the Amp Hour, there are good reasons. Customizability is one key point, and the other is that you can get them fast. We have PCB mills because we want to be able to go from design to production to testing in a day. And when you use them enough, the cost of producing your own boards with a mill is actually less than ordering from a board house. (This is especially true if we only need one board. Usually when we order from a boardhouse, we get a larger number than we need in case we have problems.)

When 3D printers are advanced enough, there is no reason they won’t be able to build complete boards with parts, including ICs, integrated into the printing process. No soldering, no reflow, better reliability, much faster…and for some people, the printers will end up paying for themselves. I think the problem is that you are underestimating how much some people will value that over, say, dodgy parts from Digikey or Maxim.

Like

4. Dave Jones - January 10, 2011

You are making invalid comparisons.
3D printers and PCB mills CANNOT be compared to chipmaker machines because EVERY design needs a custom PCB (and let’s say for arguments sake) and a custom case.
On the other hand, almost the entire electronics product design industry is built around using the HUNDREDS of THOUSANDS of PROVEN components based on PROVEN supply chains that are available. What use is a ChipMaker machine to designers when it can only produce basic crude prototype level chips?

As I’ve said there will be niches for such a machine, but they will remain very small niches.

And that’s IF the chipmaker machine is possible, which of course is borderline ridiculous pie-the-sky stuff if you are talking any form of advanced chip. 5 years is a joke, 10 years is a joke, I won’t even bother speculating further.

Who cares if you can get it quick? I can get any one of a hundred thousand PROVEN advanced parts in a day or two, sometime same day. Why the hell would I want to go to the effort to design and make my own chips? To what end? I’ll just make it out of existing parts like people have been doing for almost a century.

And that doesn’t factor in the growth of programmable digital (FPGA) and analog chips that just gets better and better each year. Forever eroding the usefulness of having a ChipMkaer machine.

And there are other levels of complexity beyond these that muddy the whole idea even further, like design tool complexity, I won’t bother going through them again.

The whole concept is so massively flawed on so many levels, that anyone who thinks it’s going to be a popular and viable way to make chips hasn’t really thought about all the industry driven practicalities of it.

Even if there was a machine for $500 (Makerbot type price), I wouldn’t bother buying one. it would be nothing but a fun toy.

As we say in Australia, “Dream’n”:

Like

mareserinitatis - January 11, 2011

I didn’t say you had to design the chips, especially since I’m not talking about a ‘chip maker’ machine…more as part of an integrated way to manufacture PCBs. I imagine that 3D printers could come loaded with specs on how to build various popular ICs. You need a 741, just tell the machine you need it, it’ll print it onto your board along with everything else. With time, the reliability of the parts could easily be on par with that of ordered parts. I think with all the focus on flexible, printed electronics, it’s very realistic to expect that the knowledge gained from creating a reliable, printed process on flexible electronics can’t help but to bleed over into 3D printing.

I think it makes far more sense for companies and individuals who are designing to invest in machines that can prototype a complete board quickly and cheaply in one step. It greatly simplifies prototyping and reving boards. This is especially true when the alternative is to order boards, order parts (hope the parts are in stock), have to solder stuff or wait for someone else to have time to put it in the pick-and-place and then reflow.

I do agree that it will probably only work for simpler ICs…some of the more sophisticated items will be too much. But I wouldn’t underestimate how convenient many people will find it to print out a multi-layer board with integrated components in just a couple hours.

Like

mareserinitatis - January 11, 2011

The other thing you’re missing is that it could be easier to buy bottles of ink to print a whole slew of chips than to buy chips individually. Also, companies would rather sell IP (in the form of a chip design) than to have to manufacture the chips themselves. Much less overhead and higher profits in selling designs rather than devices.

Like

5. chrisindallas - January 10, 2011

Lets say we wanted to fab some 40 year old technology in our garage. We’d still need equipment that could produce the same transistor, with the same electrical parameters over and over with a few microns of tolerance. Even the most expensive mechanical machines are several orders of magnitude from this tolerance, which is why photolithography has driven IC fabrication like it has.

This means we’re back to deadly chemicals, expensive equipment, etc. But lets say we can throw together something where we could make 100 transistors on one die. Why would we want to use a 40 year old process to design circuits? What circuits would you want that aren’t commercially available? How much would they cost?

Maybe Dave and I are wrong, and we’ll see fab-a-home in our life times, but I don’t see it effecting IC designers. Just give us spice models, and we can make circuits.

Like

mareserinitatis - January 11, 2011

The issues is that the smaller the scale, the larger the tolerance problem. Photolithography at the smallest scales has issues with yield. Larger parts created by a well-controlled manufacturing process are far more likely to produce consistent results where you don’t have to prove out the part every single time. Photolithography also involves a lot of nasty chemicals. There’s a lot that’s attractive in being able to print and assemble electronics without all the nasty side-effects. For instance, look at all the issues in moving away from leaded solder. If you can print a part onto your board using only copper ink (which they already do for RFID), you’ve effectively bypassed the soldering process and the problems that come from solder interconnects.

As I mentioned in my reply to Dave, I don’t anticipate people will be designing ICs from scratch (although they could if they wanted to). They’ll be using pre-loaded designs that are based on what the printer can do.

Like

6. Charles J Gervasi - January 11, 2011

I still want to know why it matters if the fab is at home. If I can order boards from Sunstone on a one-day turn with overnight delivery to Wiconsin, why do I need the fabrication to happen in my lab? If I can custom ASICs affordably it doesn’t matter if they’re made in my lab.

Like

mareserinitatis - January 11, 2011

I’m just saying that, at some point, it may be cheaper, more convenient, and easier to do it this way. Obviously that point is not now, but I wouldn’t be surprised if it’s 10 years off. One of the places my husband worked for has 3D printer-type thing that already does basic boards for proof-of-concept type work. I can only imagine that the technology for printed electronics is developed, this type of thing would be added as a feature to 3D printers. You know that businesses are going to invest in them, as will the gadget hounds. And, at some point, it may be economical for individuals to have them, as well.

Like


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: